# `imp_blend_a` — cocarry_partners@v2 **Set.** `cocarry_partners@v2` (task `cocarry@v1`) · **Stratum.** scripted · **Split.** private · **Role.** partner_arm **Identity.** `partner_id` `b6975bc340b25133` · parameter digest `66ec45f3bc06acf1ca79f5f3804aff757920c3db8c02a0cdd59675cad241c1bf` · `weights_uri` `member://cocarry_impedance/66ec45f3bc06acf1ca79f5f3804aff757920c3db8c02a0cdd59675cad241c1bf` ## What this partner does A deterministic end-effector-space PI impedance controller for one co-carry arm (ADR-026 §Decision 1): it drives its bar-end toward the cooperative goal along a straight Cartesian path through a damped pseudo-inverse Jacobian, sharing the bar load compliantly with the ego arm. `kp`/`ki` set the Cartesian PI gains, `step_max` the per-step command clip (response speed), `damping` the pseudo-inverse regularisation, and `lag_steps` a bounded transport delay (a sluggish-but-competent member: delayed but convergent tracking). ## Construction provenance Built from registry class `cocarry_impedance` (`PartnerSpec(seed=0, checkpoint_step=None)`). Private member: the exact values are **withheld** (ADR-009 as amended — published hashes, withheld parameters). They were drawn from the committed box under the maintainer-held seed; the digest below verifies any authorised local re-derivation. | parameter | committed box | value | | --- | --- | --- | | `damping` | [0.02, 0.08] | *withheld* | | `ki` | [0.4, 0.8] | *withheld* | | `kp` | [1.3, 3.8] | *withheld* | | `lag_steps` | [0, 3] (int) | *withheld* | | `step_max` | [0.02, 0.04] | *withheld* | ## Behavioural fingerprint Committed probe suite: seeds [7000, 7001, 7002, 7003] x 5 episodes against the task's reference ego (archive: `spikes/results/partner-fingerprints/cocarry_partners-v2/imp_blend_a/`, an ADR-028 v3 bundle). Summary statistics of the member's action distribution and the dyad's stress channel — enough to tell members apart without policy access (ADR-018/I3). | statistic | value | | --- | --- | | `action_abs_mean` | 0.3683 | | `action_abs_p90` | 1 | | `action_std` | 0.4705 | | `n_episodes` | 20 | | `stress_max` | 114.7 | | `stress_mean` | 94.41 | | `stress_p50` | 101.6 | | `stress_p90` | 111.8 | | `success_rate` | 1 | **Competence floor.** floor probe `fingerprint` — success 1 vs committed floor 0.75 → **PASS**